First insights into genetic diversity and relationships of European taxa of Solenopsora (Catillariaceae, Ascomycota) with implications for their delimitation

The lichen genus Solenopsora occurs predominantly in temperate and subtropical regions of the world, and the centre of diversity and distribution is in the Mediterranean, Macaronesian, and Madrean ﬂoristic regions. Taxonomic treatment of several taxa has varied over time and the concepts lack clarity. Focusing on multilocus sequence data, morphology, anatomy, chemistry, and ecological preferences, the present study investigates European Solenopsora taxa to obtain the ﬁrst insights into their genetic variation and relationships. Our results show discrepancy between the number of currently recognized taxa in Europe and the number of genetic entities identiﬁed. We recognize eight species in the genus in Europe: Solenopsora candicans , Solenopsora cesatii (including Solenopsora carpatica ), Solenopsora grisea , Solenopsora holophaea , Solenopsora liparina , Solenopsora marina , Solenopsora olivacea , and Solenopsora vulturiensis . We gathered evidence to recognize S. liparina , an edaphic vicariant of S. candicans conﬁned to ultramaﬁc rocks, as a separate species. We disclosed a previously unknown sister relationship between S. grisea and S. vulturiensis . Taxonomic synopses, geographical distribution, and an identiﬁcation key for the treated taxa are provided. Phylogenetic analyses revealed two major lineages among the European Solenopsora spp., differentiated by thallus organization, the presence/absence of rhizines on lower side of the thallus, secondary chemistry, and anatomy of upper cortex. The generic circumscription and phylogenetic position, however, appear problematic and additional studies with increased sampling including related genera are needed. © 2014 The Linnean Society of London, Botanical Journal of the Linnean Society , 2014, 176 , 203–223.


INTRODUCTION
The circumscription of species in lichen-forming fungi has largely been based on morphological and chemical characters. Lichens display few taxonomically useful characters and many of these are widely variable, and the homology of character states within and between systematic groups is difficult to assess (Printzen, 2010). Nevertheless, accurate species delimitation is crucial in biodiversity assessments and conservation biology. Molecular markers provide an invaluable tool for robust species delimitation (Lumbsch & Leavitt, 2011) but, despite advanced methods and employment of genetic data in lichen taxonomy, many lichenforming families and genera still receive little attention, resulting in conflicting taxonomies or uncertain species circumscriptions.
One of the poorly known lichen genera with conflicting taxonomy is Solenopsora A.Massal. It comprises approximately 20 species (Gilbert, Purvis & James, 2009) and predominantly occurs in temperate and subtropical regions of the world, with the exception of South America and polar regions (Ryan & Timdal, 2002). The centre of diversity and distribution is in the Mediterranean, Madrean, and South-West Australian floristic regions (Takhtajan, 1986). Despite having a relatively small genus, its thallus morphology, anatomy, and chemistry are complex.

TAXON SAMPLING
The distribution ranges of the Solenopsora taxa described from Europe and Canary Islands were assessed from herbarium specimens (BG, BP, BR, BRA, CBFS, GZU, H-NYL, O, PRA, PRC, PRM, SAV, TO, VER, W) and relevant literature sources (see Supporting information, Doc. S1). Based on these data, during 2010-2013, we sampled material representing all but three taxa (see below) reported from Europe. Each taxon was sampled from multiple sites, including type localities when possible (see Fig. 1; see also Supporting information, Doc. S2). 204 A. GUTTOVÁ ET AL.  Additional material from the USA was obtained and used for molecular analyses (herbarium MSC). Voucher specimens are deposited in SAV and O. The list of additional specimens examined only for morphology, anatomy, chemistry, and ecological requirements is indicated in the Supporting information (Doc. S3). We did not include S. caliacrae, described from Bulgaria, and S. paskovskiana, described from Ukraine, because they are known only from protologues (Cretzoiu, 1940) and the type material has not been found in relevant collections in Romania (BUC, BUCM, CL). We also did not find S. caliacrae in its type locality during our field research in 2013. We do not treat S. fumosula, described from Istria, because the species is known so far only from the type locality and we did not find it during field work. We only have the historical type material at our disposal (W). Based on the available information on the studied group, a member of Catillariaceae (Printzen, 2010), Catillaria lenticularis (Ach.) Th. Fr. (see Supporting information, Doc. S2), was chosen as the outgroup.

SELECTED LOCI, POLYMERASE CHAIN REACTION (PCR) AMPLIFICATION AND SEQUENCING
Total genomic DNA was extracted from peripheral parts of thalline lobes (Solenopsora sp. div., up to 2 mm from lobe tips), and apothecia (Catillaria lenticularis), using a DNeasy® Plant Mini Kit (Qiagen) in accordance with the manufacturer's instructions. The quality of the extracted DNA was checked on 1.5% TAEagarose gels. The quantity and purity of the isolated genomic DNA was also determined spectrophotometrically (NanoDrop 2000; Thermo Scientific). An initial screening of multiple primer combinations amplifying diverse nuclear target regions was performed using three different species. As a result, three regions of different molecular nature were selected for the subsequent analyses: the commonly applied multicopy internal transcribed spacer (ITS) region of nuclear ribosomal DNA (ITS1-5.8S-ITS2), the ketosynthase (KS) domain of the polyketide synthase gene (PKS, nonreducing fungal type I; Schmitt et al., 2005), and a single-copy protein-coding nuclear gene Mcm7 (replication licencing factor) recently shown to be phylogenetically informative (Schmitt et al., 2009;Raja et al., 2011;Truong et al., 2013). The primers used are listed in

SEQUENCE ALIGNMENT
Sequence identity was confirmed using the BLAST search tool in GenBank (Wheeler et al., 2006). Electropherograms were carefully inspected in CHROMAS LITE, version 2.01 (Technelysium Pty Ltd) for double peaks indicating heterozygous positions (Mcm7) or sequence variation among the multiple copies of the KS and ITS regions. Intra-individual polymorphic sites were coded using International Union of Pure and Applied Chemistry (IUPAC) ambiguity codes. Sequences were aligned manually in BIOEDIT, version 7.1.3.0 (Hall, 1999). Manual alignments of the KS and ITS regions were improved in GENEIOUS, version 5.6.4 (Biomatters Ltd) using MAFFT, version 6.814b (Katoh et al., 2002) plugin. A highly ambiguous indel region observed at the 5′ end of ITS1 was excluded. All sequences generated were deposited in GenBank (for accession numbers, see Supporting information, Doc. S1).

PHYLOGENETIC ANALYSIS
Phylogenetic trees were inferred using Bayesian and maximum likelihood (ML) analyses. Best-fit models of nucleotide substitutions were selected in JMODEL-  (Akaike, 1974) and implemented in the Bayesian and ML computations.
For the KS domain that included a short intron region, intron and exon sequences were defined as different partitions, specifying also the first, second, and third codon positions of the exons and unlinking model parameters across the defined partitions of the data. For the Mcm7 data that encompassed a single exon region, we also defined and unlinked three character partitions based on the first, second, and third codon positions. In the ITS dataset, two partitions corresponding to the 5.8S gene and ITS1 + ITS2 were defined (  Kumar et al., 2009). Two parallel runs with four chains of MCMC were conducted for five million generations, with a sampling frequency of every 100 generations. The settings corresponding to the best-fit evolutionary models were specified; otherwise, default priors were used. Convergence of all parameters was verified by inspecting the graphical output from the Bioportal website. The average standard deviation of split frequencies between the two simultaneous runs was checked to fall below 0.01, and the potential scale reduction factor to be close to 1.0 for each parameter. For each run, the first 5000 trees (500 000 generations) were discarded as the burn-in period, and the consensus tree was generated from the remaining trees from both runs, computing also the Bayesian posterior probabilities (BPP) for each node. ML analyses were computed in GARLI, version 2.0 (Zwickl, 2006) setting five million generations, and with multiple independent replicate runs. Searches were per-formed with random starting trees and by setting the program to stop after 20 000 generations if no improvement of the log-likelihood was detected (≤ 0.01), with a maximum of 50 000 generations. A check was made that the topologies of the resulting trees obtained from the replicate runs were the same. Branch support was assessed by 500 bootstrap replicates under the same settings as described above. Before combining the three datasets, topological congruence between the gene trees was examined using 500 replicates of ML bootstrapping under the same models described above, on each locus separately (Mason-Gamer & Kellogg, 1996). Because there was no conflict detected using a 70% reciprocal threshold (with only a slightly higher value, 74%, for a single clade in the KS gene tree), the alignments were concatenated. We conducted a Bayesian analysis (MrBayes, version 3.2.1) of a concatenated dataset of all three markers. In this case, each locus was defined as a separate partition with the rest of settings as specified above. In addition to the concatenation approach that, in some cases, may lead to poor estimation of the species tree (Kubatko & Degnan, 2007), we also employed a Bayesian coalescent-based approach to estimate a species tree using *BEAST implemented in BEAST, version 1.7.4 (Heled & Drummond, 2010). *BEAST infers the species tree directly from the sequence data and assumes that the incomplete lineage sorting/deep coalescence is the main source of inconsistency between gene trees and species trees. BEAUti, version 1.7.4, was used to generate the input file for BEAST, setting three data partitions (corresponding to the three loci), the bestfit evolutionary model for each dataset (determined in JMODELTEST, see above), assuming uncorrelated lognormal clock, a birth-death model for the species tree prior, and the remaining parameters set to default. Four independent MCMC analyses were run each for 10 million generations, sampling every 1000th generation. Convergence of all parameters to the stationary distribution in each run was assessed using TRACER, version 1.5 (Rambaut & Drummond, 2007). TREEANNOTATOR, version 1.7.4, (http:// beast.bio.ed.ac.uk/TreeAnnotator) was used to set the burn-in (discarding the first 2000 trees) and to calculate the maximum clade credibility tree.
In the concatenated and coalescent-based approaches, S. holophaea was not included (Mcm7 sequences could not be retrieved here) and S. olivacea subsp. olbiensis was represented by a single sample only (ITS sequence could not be retrieved from the sample olbiensis 2 GR). All other taxa were represented by multiple accessions (see Supporting information, Doc. S1).

MORPHOLOGICAL, ANATOMICAL, CHEMICAL, AND ECOLOGICAL ANALYSES
For all studied taxa, we scored selected morphological, chemical and ecological traits (not only in the specimens listed in the Supporting information, Doc. S1, but also for those marked in Doc. S2). We focused on thallus morphology, mode of reproduction, ascospore size, hymenium and hypothecium heights, and the presence of secondary metabolites. For ecological preferences, we studied obligate and facultative substrate and microhabitat preferences related to light and humidity. The examined features and preferences are summarized in Figure 3 and in Table 3. Microscopical examinations (light microscopy) were performed on hand-cut sections mounted in water and 10% KOH. The measurements of thallus and apothecia given in the description of the species are based on the selected examined material. Cryotome sections were prepared to study the anatomy of the cortex in detail. Chemistry was studied by thin-layer chromatography (TLC), in solvent systems A, B′ and C (Culberson & Kristinsson, 1970;Culberson, 1972;Culberson & Johnson, 1982;Orange, James & White, 2001). Microextracts of herbarium specimens with TLC determined chemistry (pannarin, zeorin, atranorin) were used as standards. Terminology regarding morphology, and anatomy of the lichen thallus follows that of Ryan, Bungartz & Nash (2002) and Ryan et al. (2012).

SEQUENCE DATA CHARACTERISTICS
Details of nucleotide alignments are summarized in Table 2, including the best-fit models of evolution selected with JMODELTEST and notes on the data partitions used. In the ITS dataset, variation in the 5.8S gene was negligible, whereas ITS1 and ITS2 were highly variable, requiring the introduction of numerous (mainly 1-or 2-bp, or longer but overlapping) indels. Separate indel scoring, however, appeared ambiguous. In the KS alignment, one short intron region with four 1-4-bp indels was identified. Two indels supported two of the main clades of the resulting gene tree (clades B and IV of A; see below) and two were singletons, each present in a single sample. The scarcity of indels contrasts here with a high number of SNPs in the alignment and implies that their phylogenetic information is negligible in this case. Thus, in both datasets, the indel characters were treated as missing data. The Mcm7 alignment included only the exon sequences with no indel variation.

GENE TREES VERSUS CONCATENATED PHYLOGENY AND COALESCENT-BASED SPECIES TREES
Bayesian phylogenetic trees based on KS, Mcm7 and ITS (see Supporting information, Figs S1, S2, S3) yielded similar topologies, showing that two species, Solenopsora holophaea (present in KS and ITS trees) and S. marina (denoted as clade B), are divergent from the rest of the species (clade A) that form four major clades (I-IV): clade I, Solenopsora grisea and S. vulturiensis (BPP KS = 0.96, BPPMcm7 = 0.99, BPPITS = 1.00); clade II, S. olivacea, including subsp. olivacea and subsp. olbiensis (BPP = 1); clade III, S. cesatii (BPP = 1); and clade IV, S. candicans, S. liparina, and two potential new species that need further study, tentatively labelled here as Solenopsora sp. 1 and Solenopsora sp. 2 (BPP = 1). Accessions of S. carpatica (indicated in bold and with an asterisk in Fig. 2; see also Supporting information, Figs S1, S2, S3) appear in the clades of S. cesatii (samples from the Western Carpathians and the Alpi Bergamasche Mts) and of S. liparina (a sample from the Sudety Mts) in all three gene trees. Resolution of the gene trees, however, was insufficient in some parts. They did not resolve the relationship between S. grisea and S. vulturiensis (clade I). The ITS tree (see Supporting information, Fig. S3) suggested an alternative branching pattern for clade II (here sister to clade I), albeit with a weak support (BPP = 0.97, bootstrap support = 52%). Similarly, the sister relationship between S. olivacea subsp. olivacea and subsp. olbiensis was supported only in the KS tree (BP = 1; see Supporting information, Fig. S1), whereas it remained unresolved in the Mcm7 and ITS trees (see Supporting information, Figs S1, S3). ML trees (not shown) displayed topologies similar to the Bayesian ones. Bootstrap support obtained from the ML analyses is depicted on the respective Bayesian trees (see Supporting information, Figs S1, S2, S3). The same two main clades (A and B) and four clades (I-IV) in clade A and relationships among them were observed for each gene tree, with topological TAXONOMY OF THE LICHEN GENUS SOLENOPSORA 209 differences regarding more terminal subclades with lower BPP support. A Bayesian phylogenetic tree inferred from the concatenated alignment showed a topology congruent with the PKS and Mcm7 gene trees in respect of the position of clade II (Fig. 2). In the coalescent-based species tree, the same four clades (all BPP = 1) and topology were obtained, except for the position of clade II that remained uncertain (Fig. 3). In the replicate runs, it was either resolved in the same position as in the concatenated tree but with low node support (BPP = 0.65) or as sister to the clade I as also seen in the ITS tree (with BPP = 0.69-1, depending on the run; Fig. 3).
Sister relationships of closely related taxon pairs S. grisea and S. vulturiensis (clade I) and S. olivacea subsp. olivacea and S. olivacea subsp. olbiensis (clade II) are confirmed in both the concatenated and species trees. The topology of the clade IV consisting of four species (S. candicans, S. liparina, Solenopsora sp. 1 and sp. 2) was congruent in the concatenated gene tree and the species tree, being identical to that in the ITS tree, and resolved the uncertainties in the KS and Mcm7 trees (see Supporting information, Figs S1, S2). Catillaria lenticularis, used as the outgroup, was resolved in the species tree in the sister position to S. marina with a high support (BPP = 1), although this relationship had no support in the concatenated gene tree (Figs 2, 3).

PREFERENCES
The studied taxa share similar dorsiventral body plans with three different thallus growth forms. In and incongruence between four independent Markov chain Monte Carlo algorithm runs: the alternative topology is indicated by grey branches with the BPP range of replicate runs being 0.69-1.00. A and B denote two main lineages, and I-IV indicate the clades referred to in the text. The species characteristics depicted on the right-hand side of the tree include: 1 thallus form: SQ -squamulose, PL -placodioid, CR -crustose; 2 secondary chemistry: PZ -pannarin and zeorin major, CA -complex of terpenoids with occasional presence of atranorin, CT -complex of terpenoids, CX -complex chemistry; 3 reproduction: X -sexual, V -vegetative; 4 habitat: C -calcareous, S -siliceous, U -ultramaphic, Mmuscicolous; 5 microhabitat: F -plane rock face, CH -chasmophyte; 6 humidity: MX -mesoxerophyte, MP -mesophyte; and 7 light: H -heliophyte to helioscyophyte, O -ombrophyte, OS ombro-scyophyte, HS -helioscyophyte. 212 A. GUTTOVÁ ET AL.
Pruina is present in the taxa of both main clades. It might cover the entire thalline surface (e.g. S. candicans, S. cesatii, S. sp. 1) or it can be limited to the thalline margins (e.g. S. marina, S. vulturiensis, S. liparina, S. sp. 2). We noted differences in pruina cover in S. grisea. Thalli on rocks more exposed to light were whitish on their entire surfaces, whereas, on shaded rocks (in woodlands), the pruina was limited to lobe ends. This feature might have been a source of misidentifications of S. grisea in the past, especially when a small, marginal piece was available for closer examination.
The taxa in the B clade are characterized by a well developed upper cortex, either paraplectenchymatous or composed of anticlinally oriented, gelatinized hyphae, and sexual reproduction. Clade A comprises taxa with upper cortex of irregularly interwoven hyphae with a continuous epinecral layer, and both sexual and vegetative modes of reproduction (blastidia, soralia). Both reproduction modes occur in S. grisea, S. vulturiensis, and S. olivacea subsp. olbiensis (the type species forming apothecia and soralia; Nylander, 1876; H-NYL29314).
Clade B includes samples containing unknown terpenoids (products of the mevalonic acid pathway). In clade A, we found a combination of polyketide pathway products (pannarin, atranorin) and mevalonic acid pathway products (zeorin, unknown terpenoids).
The characters described above are summarized in Table 3 and depicted on the species tree (Fig. 3). It is apparent that characters such as squamulose thallus organization, pruina on the thalline surface, ascospore size, hymenium and hypothecium height, and complex secondary chemistry are homoplasious.
Each Solenopsora taxa has peculiar, unique microhabitat preferences (e.g. light, humidity), thus occupying specific microniches in the same outcrop. They are mostly saxicolous, growing on calcareous sedimentary rocks, such as limestone, dolomite, and conglomerate, or on siliceous igneous rocks, such as basalt or ultramafic rocks. With respect to ecological requirements and distribution, we can divide them into two groups. The first group includes the species with narrow ecological amplitudes confined to the Mediterranean climate: S. grisea, S. marina, S. olivacea subsp. olivacea, and S. olivacea subsp. olbiensis. The second group comprises the taxa with wider ecological amplitudes and distribution ranges: S. candicans, S. cesatii, S. vulturiensis, and S. holophaea.
Two taxa (S. liparina and S. sp. 2) feature a particular position because they are confined to ultramafic rocks and appear as edaphic vicariants of S. candicans that typically grows on calcareous rocks. These three taxa are part of clade IV (Fig. 3).

MULTILOCUS APPROACH
Fungi in general, unlike other pluricellular organisms, have a relatively limited set of taxonomically useful morphological traits, and the specific genetic control underlying most diagnostic characters is currently unknown . Furthermore, as a result of the small size of the fungal chromosomes, there are no karyological data that are significant in the taxonomy of vascular plants (Crespo & Pérez-Ortega, 2009).
Molecular data provide an excellent basis for the critical evaluation of phenotypic characters, and help to trace previously overlooked differences (Lumbsch & Huhndorf, 2007). They are also extremely useful for assessing whether the investigated traits are reliable or show homoplasy (Fontaine, Ahti & Piercey-Normore, 2010;Muggia et al., 2011) and revealing cases of cryptic speciation (Schmitt et al., 2005;Vondrák et al., 2009). The present study is largely based on inferences from DNA sequences of three tested nuclear regions. Morphological and chemical data were also gathered and interpreted in the light of the presented genetic data. Both the gene tree and coalescent-based species tree approaches were employed to infer a robust phylogenetic inference of the studied taxa. The ITS region is the most popular genetic marker for mycological systematics at low taxonomic levels and has been used for barcoding (Nilsson et al., 2012). This is mainly because of straightforward amplification, even with partly or severely degraded DNA, and the availability of universal primers. Phylogenetic inference based on ITS can, however, be biased and misled by the orthology/ paralogy conflation (e.g. ITS heterogeneity reported from Physcia aipolia (Ehrh. ex Humb.) Fürnr. and Physcia stellaris (L.) Nyl.; Simon et al., 2005), intraand interlocus sequence homogenization, a higher level of homoplasy, and alignment uncertainty mostly as a result of indel accumulation (Álvarez & Wendel, 2003). Alignments with low confidence have posed problems especially among representatives of higher taxonomic categories (Peršoh, Beck & Rambold, 2004), which we also encountered during the present study, when the highly ambiguous 5′ end of the ITS1 had to be excluded. The exclusive use of ITS has therefore been abandoned, and it has been suggested that more robust insights can be achieved by a TAXONOMY OF THE LICHEN GENUS SOLENOPSORA 213 multilocus approach including also single-or low-copy nuclear genes (Álvarez & Wendel, 2003, Hofstetter et al., 2007Truong et al., 2013). Multicopy nuclear PKS (KS) genes have been shown to be reliable markers (Schmitt et al., 2005). The KS region, although less variable than ITS, has proved useful in some phylogenetic studies (e.g. on species of Lecanora: Grube & Blaha, 2003;and the Cladonia gracilis group: Fontaine et al., 2010). Mcm7 is a single-copy, protein-coding nuclear gene, recently recognized as a useful marker to infer higher and lower level taxonomic relationships (Raja et al., 2011;Truong et al., 2013). Although it may be more difficult to amplify by PCR, it can provide a better phylogenetic resolution than the other markers (e.g. ITS; Truong et al., 2013). Our results show that KS and Mcm7-based trees provided largely congruent branching patterns with minor discrepancies in the topology and resolution in more terminal clades only, whereas the ITS-based tree suggested a certain discordance in a different placement of the clade II. Importantly, both the concatenated gene tree and the species tree provided more resolution, and also resolved the sister positions of two pairs of closely-related taxa (S. olivacea subsp. olivacea and subsp. olbiensis, clade II; S. grisea and S. vulturiensis, clade I), which were inconclusive in the individual gene trees. However, the position of clade II (S. olivacea) remained uncertain, and the source of this conflict needs to be finally resolved by additional sampling or markers (other nuclear low copy or mitochondrial regions). In the present study, the inferred phylogenetic trees reveal two divergent lineages (clades A and B) among European representatives of Solenopsora. The two lineages exhibited highly divergent sequences for all three markers and appeared as two well-supported clades. This division is also supported by several diagnostic features, such as thallus organization, rhizines on lower side of the thallus, secondary chemistry, and anatomy of upper cortex (Fig. 3). At this point, we consider that it would be premature to outline firm phylogenetic or taxonomic conclusions regarding the genus circumscription because our focus was on the European taxa only. Related genera and Solenopsora taxa from the other parts of the world need to be included to revise the generic concept of the studied group; for example, Solenop-sora chihuahuana B.D.Ryan & Timdal, Solenopsora cladonioides B.D.Ryan & Timdal, Solenopsora crenata (Herre) Zahlbr., and Solenopsora cyathiformis (Szatala) Van den Boom from North America (Ryan & Timdal 2002 and Solenopsora elixiana Verdon & Rambold, Solenopsora sordida (C.W.Dodge) D.J.Galloway, and S. tasmanica Kantvilas from the Southern Hemisphere (Galloway, 2004;Kantvilas, 2004;Van Den Boom et al., 2011).

RELATIONSHIPS AMONG THE TREATED
In the lineage A, clade I is resolved as the firstbranching; it includes S. grisea and S. vulturiensis (Fig. 3). Although the individual gene trees did not resolve their relationships, the concatenated and species trees confirm here for the first time the sister position of these two taxa. Among the synapomorphies that these species share the presence of both sexual and vegetative (blastidia) reproduction modes and complex chemistry. The species differ in thallus morphology and size (see key to the treated Solenopsora taxa), substrate preferences (calcareous versus siliceous rocks), and distributional range [strictly Mediterranean climate (S. grisea) versus wider occurrence to temperate to boreal zones of the Northern and Southern Hemispheres (S. vulturiensis; see Supporting information, Doc. S3)].
Our results suggest that the crustose species S. olivacea (clade II) includes two infraspecific lineages, which we consider subspecies rather than varieties (McNeill et al., 2012: chapter 1, art. 4). They differ in reproduction mode, with the nominate subspecies olivacea reproducing sexually and subspecies olbiensis obligately reproducing vegetatively via soredia (apothecia are exceptionally recorded, cf. Nylander, 1876). The two subspecies also differ in ecological preferences: S. olivacea subsp. olivacea grows on rock faces in shaded and open habitats, whereas S. olivacea subsp. olbiensis is an obligatory chasmophyte; if it covers perpendicular rock faces, then these are situated in humid and shaded forests.
The S. cesatii clade (clade III) is well supported in all the analyses, although it also includes the samples so far assigned to the taxon S. carpatica from the Western Carpathians (Slovakia) and the Alpi Bergamasche (Italy) (see Supporting information, Figs S1, S2, S3). These results confirm Kotlov's concept (Kotlov, 2004), that S. carpatica is a synonym of S. cesatii (see Supporting information, Table S1). The revealed variation patterns do not favour recognizing infraspecific taxa within S. cesatii.
Clade IV includes two well-delimited species: S. candicans and S. liparina (Figs 2, 3). Solenopsora liparina also includes a sample until now identified as S. carpatica from the Sudety Mts (Czech Republic). Our data do not support the systematic position of S. liparina as a variety of S. cesatii [S. cesatii var. olivacea (Bagl.) Kotlov;Kotlov, 2004] (see Supporting information, Table S1). The phylogenetic analyses disclosed also two separate, so far unrecognized/ unindentified entities. Solenopsora sp. 1 was collected from perpendicular calcareous rock faces in Turkey. Solenopsora sp. 2 grows on silica-carbonate rocks in California, USA (Rajakaruna et al., 2012). We consider it premature at this stage to describe these two new species formally because more material would be desirable to sample and analyze, considering also the other, non-European taxa of the genus.

MORPHOLOGICAL, REPRODUCTIVE, AND CHEMICAL TRAITS
Lichen genera are often delimited on the basis of thallus growth form, which is a highly variable character (Printzen, 2010). Several characters, thalline form including, that are homoplasious at a certain level (e.g. family) can be differential or synapomorphic at lower levels such as genera or species (Crespo & Pérez-Ortega, 2009). The point is that these forms, having particular anatomical structure and function (Grube & Hawksworth, 2007), are optimal to the successful colonization of particular ecological niches. Foliose-squamulose thallus is present in lineage B, with two occurrences recorded also in lineage A (Solenopsora sp. 2 and S. vulturiensis). The squamules/ folioles of the taxa in lineage B are partly detached from the substrate, corticate on the upper surface (cortex well developed, either paraplectenchymatous or of anticlinally oriented, gelatinized hyphae) and with rhizines on lower surface. Variation in the morphology and anatomy of the lichen thallus has been shown to control water storage capacity and evaporative resistence (Pintado, Valladares & Sancho, 1997). Well developed upper cortex promotes thallus hydrated for longer period (Fos et al., 1999) and limits CO 2 exchange necessary for photosynthesis and respiration (Green, Snelgar & Brown, 1981). Development of rhizines was shown to be a mechanism for increasing water storage capacity in Parmelina pastillifera (Harm.) Hale (Tretiach & Brown, 1995). Grube & Hawksworth (2007) raised the hypothesis that the cortex imposes a physical necessity to let the thallus grow as a detached form. These traits, in case of S. holophaea and S. sp. 2 (cluster IV), may help to keep the thallus hydrated longer and decrease too high irradiation because these species grow in open habitats. Solenopsora marina, however, is confined to shaded and humid sites. Non-rhizinate placodioid thallus of brittle areoles and lobes attached to the substrate dominate in lineage A clades I, III, and IV. Clade II includes taxa with crustose thallus (s.s., not forming lobes). The studied entities demonstrate that various types of thallus categories may occur in closely related taxa, as is the case of S. grisea (placodioid thallus) and S. vulturiensis (crustose-squamulose thallus).
Pruina is a concept widely used in lichenology. It includes many types of surface coverings (e.g. calcium oxalate, anthraquinones, dead mycobiont tissues) and its occurrence has been applied as an important species character in some genera (Timdal, 1984;Heiðmarsson, 1996;Wei & Wei, 2012). Different hypotheses about the ecological significance have been summarized by Wadsten & Moberg (1985) and Heiðmarsson (1996). Giordani, Modenesi & Tretiach (2003) provided experimental evidence that formation of calcium oxalate minerals in lichens is at least partially biologically controlled. Metabolic capability of forming calcium oxalate may be lost during thallus development (pruina limited to the younger parts of thalli), or retained (pruina covering entire thalline surface). Disappearance/dissolution of calcium oxalate as a result of its metastability in humid and warm habitats, creating different light conditions for the bionts, may allow the formation of soredia in mature thalline parts (Modenesi et al., 2001). Solenopsora taxa exhibit characteristic traits regarding surface coverings. The presence of pruina appears homoplastic in the studied group and suggests that we may be dealing with an ecologically plastic character, which should be interpreted with caution.
Both reproduction modes (sexual and vegetative) occur in the studied group, with two apparent complementary functions: long-distance dispersal of potentially novel genotypes via sexually derived ascospores and rapid spread of locally-adapted genotypes via asexual propagules (Seymour, Crittenden & Dyer, 2005). The taxa in lineage B are characterized by sexual reproduction. Lineage A comprises taxa with both sexual and vegetative modes of reproduction. We observe a primarily vegetative reproduction mode in S. grisea and S. vulturiensis (blastidia or soralia like structures) and in S. olivacea subsp. olbiensis (soralia), with occasionally occurring apothecia. Ascospore morphology is frequently a key character to distinguish genera in lichens. Molecular evidence indicates that most types have evolved several times independently in Lecanoromycetes and constitute a highly homoplasious character (Printzen, 2010). Members of Solenopsora have one-septate, ellipsoid ascospores, with subtle variation in the ascospore apex shape: acute-acuminate, obtuse, pointed (Verdon & Rambold, 1998;Kantvilas, 2004). Sporesize and shape did not segregate the studied taxa into certain groups (Table 3). Moreover, similar to the observations of Kantvilas (2004), the studied fertile specimens yielded few, mature, well-developed ascospores, therefore making a detailed assessment of spore shape difficult (Kantvilas, 2004). Further study is also needed to explore this element in TAXONOMY OF THE LICHEN GENUS SOLENOPSORA 215 Solenopsora and to clarify whether spores have reproductive function.
The widespread occurrence of most secondary metabolites across the lichenized ascomycetes indicates that they convey little information on phylogenetic relationships, although some have proved useful when distinguishing species or even at higher levels (Printzen, 2010) or at least suggest general trends . Secondary chemistry is one of the characters distinguishing the two main lineages identified in the present study. Lineage B features complex secondary chemistry of accessory type compounds (terpenoids, unknown substances), whereas the taxa in lineage A produce either chemosyndromes with pannarin and zeorin as major substances, or accessory type compounds (terpenoids, unknown substances) with occasional records of atranorin, as in S. grisea (Table 3). Moreover, other Solenopsora spp. not included in the present study produce peculiar major compounds [lobaric acid in S. elixiana (Verdon & Rambold, 1998;Van Den Boom et al., 2011); brialmontin in S. tasmanica (Kantvilas, 2004); and gangaleoidin in S. cyathiformis (Van den Boom & Ryan, 2004)]. Additionally, S. chihuahuana, S. cladonioides, and S. crenata produce pannarin and zeorin as major substances (Ryan & Timdal, 2002. The phylogenetic position of these taxa and the evolutionary significance of their secondary compounds need further exploration.

CONSIDERATIONS
Lichens are broad-ranging organisms and are sensitive to air humidity. They can contribute to outline the bioclimatic features of an area (Nimis & Losi, 1984). They often have large geographical ranges, with many species showing high potential for long-distance dispersal, also across strong physical barriers (mountains, ocean, sea). One of the most striking distribution models found in lichens is the biogeographical Mediterranean pattern, in which several species are distributed across different regions with a Mediterranean climate on different continents (Crespo & Pérez-Ortega, 2009). Although the total distribution range of all Solenopsora spp. is still not sufficiently explored, and for some species only scarce records exist (e.g. S. chihuahuana, S. crenata, S. cladonioides), the present study indicates that the treated taxa are a good example of this kind of distribution. The taxa studied here are successful and widespread in the circum-Mediterranean and Atlantic Europe, whereas their occurrence is limited in continental parts of Europe. According to the data available until now, we can characterize four taxa as strictly Mediterranean, reaching the Canary Islands: S. grisea, S. marina, S. olivacea subsp. olivacea, and S. olivacea subsp. olbiensis. Siliceous S. vulturiensis grows along the Atlantic coast of Portugal, Great Britain and reaches southwest Norway (Jørgensen & Nordin, 2009). We include here a recent outlying finding in western Iceland. Solenopsora candicans and S. cesatii were able to disperse to more continental habitats in Europe with oro-Mediterranean conditions. However, additional collecting is necessary in the Carpathians and Alps (where they frequently grow on places that are difficult to access) to complement the picture of the distribution of these species in Europe.
Solenopsora taxa are mostly saxicolous, growing on calcareous sedimentary rocks or siliceous igneous rocks. Several lichens collected from ultramafic (serpentinite) substrates in Europe have been described as being new to science, even though it is unclear whether these are truly ultramafic endemics or species that are rare and were collected only from those substrates (Favero-Longo, Isocrono & Piervittori, 2004). Our results show that S. liparina and S. sp. 2 growing on serpentinites form two genetically distinct and well-supported entities. This supports observations that the stressful edaphic conditions of ultramafic substrates can promote speciation and the evolution of ultramafic endemism (Rajakaruna et al., 2012).

CONCLUSIONS
In the present study, multilocus DNA sequence data revealed two major lineages that are differentiated by thallus organization, rhizines on lower side of the thallus, secondary chemistry, and anatomy of the upper cortex. Our results clarified the relationships between the taxa treated under Solenopsora occurring in Europe. We recognize eight species in Europe: S. candicans, S. cesatii, S. grisea, S. holophaea, S. liparina, S. marina, S. olivacea, and S. vulturiensis. We showed that there is no support for recognition of S. carpatica; the collections from the Western Carpathians, including the type, and from the Alpi Bergamasche represent the species S. cesatii. The name S. carpatica thus falls into synonymy of S. cesatii. Our results confirm the concept of Kotlov (2004) proposing that S. carpatica is a synonym of S. cesatii and, provide supporting molecular, morphological, chemical, and ecological arguments, which were missing so far. Solenopsora cesatii is not a complex of infraspecific taxa, as it was treated in the past (see Supporting information, Table S1). It represents a well-delimited, genetically homogeneous entity. The collection from the Sudety Mts was grouped with the specimens representing S. liparina. We gathered molecular, morphological, chemical, and ecological evidence to recognize S. liparina as a sepa-rate species, being an edaphic vicariant of S. candicans, confined to ultramafic rocks. We disclosed so far unknown sister relationship between S. grisea and S. vulturiensis. Moreover, we disclosed two new entities, Solenopsora sp. 1 collected from perpendicular calcareous rock faces in Turkey as a sister taxon to S. candicans, and Solenopsora sp. 2 collected on silica-carbonate rocks in California (Rajakaruna et al., 2012), being a closest relative to S. liparina. However, we consider it premature at this stage to

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article at the publisher's web-site: Figure S1. Bayesian majority-rule consensus tree with mean branch lengths based on the ketosynthase (KS) domain of PKS genes. The posterior probability values are indicated above the branches (before slash); thick branches indicate high support (> 0.90). Italicized values (after slash) show bootstrap support > 50% assessed by maximum likelihood analyses. I-IV denote the main clades referred to in the text. The codes of accessions follow Doc. S1. The codes marked in bold and accompanied by asterisk indicate the accessions classified as Solenopsora carpatica. Figure S2. Bayesian majority-rule consensus tree with mean branch lengths based on the Mcm7 gene. The posterior probability values are indicated above the branches (before slash); thick branches indicate high support (> 0.90). Italicized values (after slash) show bootstrap support > 50% assessed by maximum likelihood analyses. I-IV denote the main clades referred to in the text. The codes of accessions follow Doc. S1. The codes marked in bold and accompanied by asterisk indicate the accessions classified as Solenopsora carpatica. Figure S3. Bayesian majority-rule consensus tree with mean branch lengths based on the ITS (ITS1-5.8S-ITS2) region of nuclear ribosomal DNA. The posterior probability values are indicated above the branches (before slash); thick branches indicate high support (> 0.90). Italicized values (after slash) show bootstrap support > 0.50 assessed by maximum likelihood analyses. I-IV denote the main clades referred to in the text. The codes of accessions follow Doc. S1. The codes marked in bold and accompanied by asterisk indicate the accessions classified as Solenopsora carpatica. Table S1. Overview of different taxonomic concepts applied in genus Solenopsora over time. Document S1. List of specimens sequenced in the present study. Species name, voucher information [geographical location, collector(s), collection number, herbarium acronym] and GenBank accession numbers for ITS, KS, and Mcm7 regions are indicated, respectively. A dash (-) denotes missing information. Document S2. List of additional specimens examined for morphology, anatomy, chemistry (*), and ecological requirements. Document S3. Synopsis of the European Solenopsora taxa.
TAXONOMY OF THE LICHEN GENUS SOLENOPSORA 223